Dear developers!
In earlier versions of ALIGNN it was easy to understand the history of progress that was in format like
{"loss": [0.5624889135360718, 0.46717020869255066, 0.5194703340530396,.... "mae": [0.4735216498374939, 0.4803052544593811, 0.48574551939964294, 0.4267691969871521...
Now it writes something like
[[2.679829239845276, 2.679829239845276, 0, 0, 0, 0], [1.4591001272201538, 1.4591001272201538, 0, 0, 0, 0], [1.637611210346222, 1.637611210346222, 0, 0, 0, 0]]
which I don't quite understand.
In addition, why in the val_results.json there are two values in square brackets like
[{"id": "POSCAR-JVASP-64240.vasp", "target_out": [0.0, 0.0], "pred_out": [-0.6871086359024048, 0.9857781529426575], "target_atomwise_pred": [], "pred_atomwise_pred": [], "target_grad": [], "pred_grad": [], "target_stress": [], "pred_stress": []}, {"id": "POSCAR-JVASP-107772.vasp", "target_out": [0.0, 1.569000005722046], "pred_out": [-0.8550516963005066, 1.1787383556365967], "target_atomwise_pred": [], "pred_atomwise_pred": [], "target_grad": [], "pred_grad": [], "target_stress": [], "pred_stress": []}]
even for a single-property regression task?
Dear developers!
In earlier versions of ALIGNN it was easy to understand the history of progress that was in format like
{"loss": [0.5624889135360718, 0.46717020869255066, 0.5194703340530396,.... "mae": [0.4735216498374939, 0.4803052544593811, 0.48574551939964294, 0.4267691969871521...Now it writes something like
[[2.679829239845276, 2.679829239845276, 0, 0, 0, 0], [1.4591001272201538, 1.4591001272201538, 0, 0, 0, 0], [1.637611210346222, 1.637611210346222, 0, 0, 0, 0]]which I don't quite understand.
In addition, why in the val_results.json there are two values in square brackets like
[{"id": "POSCAR-JVASP-64240.vasp", "target_out": [0.0, 0.0], "pred_out": [-0.6871086359024048, 0.9857781529426575], "target_atomwise_pred": [], "pred_atomwise_pred": [], "target_grad": [], "pred_grad": [], "target_stress": [], "pred_stress": []}, {"id": "POSCAR-JVASP-107772.vasp", "target_out": [0.0, 1.569000005722046], "pred_out": [-0.8550516963005066, 1.1787383556365967], "target_atomwise_pred": [], "pred_atomwise_pred": [], "target_grad": [], "pred_grad": [], "target_stress": [], "pred_stress": []}]even for a single-property regression task?